diff --git a/scripts/backfill_classifications.py b/scripts/backfill_classifications.py
index 98d7b8d..ccb210a 100644
--- a/scripts/backfill_classifications.py
+++ b/scripts/backfill_classifications.py
@@ -19,7 +19,7 @@
 print("CLASSIFICATIONS backfill BEFORE Time =", current_time)
 
 with psycopg.connect(f"host={PANOPTES_CONN} port={PANOPTES_PORT} dbname={PANOPTES_DB} user={PANOPTES_USER} password={PANOPTES_PW} sslmode=require") as panoptes_db_conn, psycopg.connect(f"host={TIMESCALE_CONNECTION} port={TIMESCALE_PORT} dbname={ERAS_DB} user={ERAS_USER} password={ERAS_PW} sslmode=require") as timescale_db_conn:
-    with panoptes_db_conn.cursor(name="panoptes_cursor").copy("COPY (select id as classification_id, created_at as event_time, updated_at as classification_updated_at, TO_TIMESTAMP(metadata ->> 'started_at', 'YYYY-MM-DD HH24:MI:SS') as started_at, TO_TIMESTAMP(metadata ->> 'finished_at', 'YYYY-MM-DD HH24:MI:SS') as finished_at, project_id, workflow_id, user_id, string_to_array(replace(replace(replace(metadata ->> 'user_group_ids', '[', ''), ']', ''), ' ', '' ), ',')::int[] as user_group_ids, EXTRACT(EPOCH FROM TO_TIMESTAMP(metadata ->> 'finished_at', 'YYYY-MM-DD HH24:MI:SS') - TO_TIMESTAMP(metadata ->> 'started_at', 'YYYY-MM-DD HH24:MI:SS')) as session_time, created_at, updated_at from classifications where id < %s) TO STDOUT (FORMAT BINARY)", (FIRST_INGESTED_CLASSIFICATION_ID,)) as panoptes_copy:
+    with panoptes_db_conn.cursor(name="panoptes_cursor").copy("COPY (select id::bigint as classification_id, created_at as event_time, updated_at as classification_updated_at, CASE WHEN metadata ->> 'started_at' ~'\d{1,2}\/\d{1,2}\/\d{2,4}' THEN to_timestamp(metadata ->> 'started_at', 'MM/DD/YYYY HH24:MI') ELSE TO_TIMESTAMP(metadata ->> 'started_at', 'YYYY-MM-DD HH24:MI:SS') END started_at,  CASE WHEN metadata ->> 'finished_at' ~'\d{1,2}\/\d{1,2}\/\d{2,4}' THEN to_timestamp(metadata ->> 'finished_at', 'MM/DD/YYYY HH24:MI') ELSE TO_TIMESTAMP(metadata ->> 'finished_at', 'YYYY-MM-DD HH24:MI:SS') END finished_at, project_id::bigint, workflow_id::bigint, user_id::bigint, string_to_array(replace(replace(replace(metadata ->> 'user_group_ids', '[', ''), ']', ''), ' ', '' ), ',')::int[] as user_group_ids, EXTRACT(EPOCH FROM (CASE WHEN metadata ->> 'finished_at' ~'\d{1,2}\/\d{1,2}\/\d{2,4}' THEN to_timestamp(metadata ->> 'finished_at', 'MM/DD/YYYY HH24:MI') ELSE TO_TIMESTAMP(metadata ->> 'finished_at', 'YYYY-MM-DD HH24:MI:SS') END) - (CASE WHEN metadata ->> 'started_at' ~'\d{1,2}\/\d{1,2}\/\d{2,4}' THEN to_timestamp(metadata ->> 'started_at', 'MM/DD/YYYY HH24:MI') ELSE TO_TIMESTAMP(metadata ->> 'started_at', 'YYYY-MM-DD HH24:MI:SS') END)) as session_time, created_at, updated_at from classifications where id < %s order by id) TO STDOUT (FORMAT BINARY)", (FIRST_INGESTED_CLASSIFICATION_ID,)) as panoptes_copy:
         with timescale_db_conn.cursor().copy("COPY classification_events FROM STDIN (FORMAT BINARY)") as timescale_copy:
             for data in panoptes_copy:
                 timescale_copy.write(data)